1. Field of the Invention
The present invention relates to a charged particle beam exposure method and a charged particle beam exposure device, and more particularly, to a method for creating exposure data for exposing patterns on a semiconductor wafer by means of a charged particle beam, such as an electron beam exposure, and a charged particle beam exposure device for implementing this method.
2. Description of the Related Art
The a charged particle beam exposure, such as an electron beam, is able to expose patterns of the sub-micron order and is used in the fabrication of highly integrated LSIs. In particular, recently, in addition to its use in forming masks, methods whereby the charged particle beam is irradiated directly onto the resist formed on a semiconductor wafer have been used.
In the design stage for LSIs, pattern data for a multi-layered structure is created in order to form a desired integrated circuit. The resist on a semiconductor wafer or the resist on a mask substrate is exposed according to this pattern data. The resist is subject to a chemical reaction generated by the energy of a charged particle beam which is irradiated onto the resist layer.
In this case, it is important to take note of the proximity exposure effect, which is caused by forward or backward scattering of the beam in the resist when the charged particle beam is irradiated onto the resist. The proximity exposure effect is a phenomenon which causes the energy of a charged particle beam irradiated onto a particular region to spread into adjacent regions due to scattering of the beam. For example, in a region where the exposure pattern density is high, after developing, a pattern may have broadened due to the effect of beam energy from a charged particle beam irradiated onto an adjacent exposure pattern region.
Alternatively, in a region where the exposure pattern density is low, there will be no effect due to energy from adjacent regions, and the pattern after developing may be condensed or narrow.
Therefore, the designed exposure data must be corrected, taking this proximity exposure effect into consideration. The present applicants have proposed such a method for correcting exposure data in Japanese Patent Application 8-13354 (Japanese Unexamined Patent 8-321462), dated 29.sup.th January 1996.
Briefly stated, in the method proposed in this patent application, a plurality of mask areas are generated in a sub-field, the pattern density in each of these areas is corrected in accordance with the effects due to the pattern density in surrounding areas, an effective pattern density taking the proximity exposure effect into account is determined, the quantities of exposure (which means beam intensity exposure time, herein after quantity of exposure) for the patterns in each area are revised in accordance ith this effective pattern density, and supplementary exposure patterns are generated additionally.
However, since the supplementary exposure patterns are produced by generating areas of a particular size without relation to the position of patterns located in a sub-field, and by taking these areas as the pattern units, there are cases where it is not possible to generate a suitable supplementary exposure pattern corresponding to an actual exposure pattern. Moreover, if the pattern density in an area is corrected with regard to the effect of the pattern density in surrounding areas, in some cases, the distance between the areas may be different from the distance between the actual pattern groupings, and in this event, it is not possible to account for the proximity exposure effect accurately. Furthermore, if the aforementioned areas are generated, the pattern density is corrected, and a supplementary exposure pattern is produced, separately in a similar manner, for each of a plurality of sub-fields which have the same exposure pattern and are positioned by repetition of, then this lengthens the data processing step for no purpose, and is not suitable for creating exposure data for highly integrated LSIs.
Furthermore, when using charged particle beam exposure, the throughput in variable square beam exposure systems deteriorates as the number of patterns increases, and therefore exposure by means of a block mask which is broader than the variable rectangle and comprises a plurality of patterns is used for regions wherein the same pattern is formed repeatedly. Since a relatively broad region comprising a plurality of patterns can be exposed in a single beam irradiation cycle, throughput can be increased.
If a block mask is used, then a plurality of patterns are comprised within a block mask and there may be respective differences between the line widths in these patterns. However, in block mask exposure, all of the patterns contained therein must be exposed with the same quantity of beam. According to the method for generating exposure data described above, the quantity of exposure is set depending on the pattern shape. In particular, the quantity of exposure is reduced when the line width is thick, and it is increased when the developed line width is narrow. This is because, when the line width is thick, the actual quantity of exposure is increased due to the proximity exposure effect from surrounding regions, and the developed line width becomes thicker, whereas when the line width is narrow, the developed one becomes narrower. Accordingly, when a block mask is used, the minimum line width is detected, for example, and the quantity of exposure is set according to this line width.
In this case, if there is a pattern with a narrow line width in a portion of the block mask, there may be cases where the quantity of exposure is set to an excessively high value accordingly, and the pattern width after developing will become too thick.
A problem arising when using block masks is that since there is a high pattern density in the block mask, when beam exposure is conducted repeatedly using a block mask, there is a tendency for the developed patterns to become thicker, due to the effect of "Coulomb interaction". This effect is especially notable at the peripheral edges of regions which are exposed repeatedly using a block mask.